690 likes | 715 Views
Reducing uncertainty in the prediction of global warming some pesky cloud obstacles. Brian Mapes doubting reductionist University of Miami. Sources. Radiation: Robin Hogan, ECMWF Ann. Seminar Sep 2008 available on web: presentation and writeup
E N D
Reducing uncertainty in the prediction of global warming some pesky cloud obstacles Brian Mapes doubting reductionist University of Miami
Sources • Radiation: • Robin Hogan, ECMWF Ann. Seminar Sep 2008 • available on web: presentation and writeup • Consults with live-in radiation guru P. Zuidema • Cloud feedbacks: • Largely from reading list • J. Clim. reviews by Stephens 2005 and Bony 2006 • Email correspondences and conversations • Bruce Weilicki, NASA Langley (ERB matters) • Larry DiGiralmo, Illinois (some scale issues) • Brian Soden, Miami
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Integrals: triumphs of atm. RT physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs to be relatable to above • Tuning • compensating errors (better than some other kind!) • any help for sensitivity? • Prospects for understanding cloud changes • in models: so what? • analogues in observations? • via conceptualizations ... a lot is being learned even as uncertainty fails to shrink
Univariate conceptual model from Stephens review - critique • The system comprises a whole lot of things • Global mean Ts well defined but how meaningful? • What is the phys/phil status of a math. average? • e.g. can be acausal (instant across space, nonlocal in time, etc.) • Relevant for interpreting what parts of DQ are “F(DTs)”
Feedbacks and sensitivity • ‘base’ negative feedback: ~ -3.2 W m-2 per K • Largely Planck feedback • -3.8 = d/dT(sT4) at global Teff = 255K • Sensitivity a 1/(Sfeedbacks -3.2) • 0 unstable climate (infinite sensitivity)
Runaway warming!! 3.2 Bony et al. 2006
cloud changes cause warmer world to emit less brighter in warmer world darker SW and LW cloud feedback Net cloud feedback from 1%/ yr CMIP3/AR4 simulations courtesy of I Held who credits B. Soden
multimodel net cloud feedback Soden Held... 2008 where d cloud causes less emission, darker albedo, or both changes in cloudiness • IPCC ch 10 • The mid-level mid-latitude decreases are very consistent, amounting to as much as one-fifth of the average cloud fraction simulated for 1980 to 1999. • Much of the low and middle latitudes experience a decrease in cloud cover, simulated with some consistency. There are a few low-latitude regions of increase, as well as substantial increases at high latitudes.
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs to be constrainable from above • Tuning = right mean answer for nonright reasons • Prospects for understanding cloud changes • in ensembles of runs of ensembles of GCMs • via conceptualizations • a lot is being learned as uncertainty fails to shrink
Radiation budget: a vast integral • global warming = • [TOArad] = ∫∫dfdl ∫dn ∫dz ∫∫dW [R]
knowns • complete integral ===~ 0 over long integration times • and presumably in preindustrial Holocene • must be maintained by overall negative feedback • Planck still king • Cleanly separable into compensating LW and SW halves, each 235 Wm-2 in global mean • equal and opposite • depend on planetary albedo and Temis • quiz: which was/is easier to guess/ measure?
Blankly: how do we compute an integral? • Reductionist extreme: • Model the integrand R explicitly and precisely • From fundamental physics • right values, for right reasons • (so sensitivity to perturbations is right too) • Holistic extreme: • Go measure the answer • or more importantly for GW, changes of the integral for a known perturbation of the integrand(forcing) • Wise: a mixed approach • Model the integrand, but by broad based estimation • physics, but also empiricism wherever can • bracket uncertainties • final accuracy depends on chain of judiciousness • “uncertainty” is both physical and social
Steps in integrating Maxwell eqs. z up to TOA (overlap) seasons, ENSO... small-scale structure geospace (lat, lon) particle ensemble angle integral wavelength integral micro meso macro subgrid schemes GCM grid sums atmospheric radiation physics
The Scale Problem “macro-” and “micro-” (physics, economics, etc.) • both intellectually on firm ground, if hard to reconcile • Micro: • Basic units obey locallaws of interaction • physics: “air parcel” jostlings, thermo • humanities: human nature, drives, responses to stimuli • Macro: • Whole system constrained by integrallaws of conservation • physics: conservation of mass, energy, momentum • humanity: demographics (fertility, nutrition, etc.). • Meso: in between: vast, important, but mushy • Only statistics... are they laws, or just descriptions?
Yechh – take me back to physics • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs to be constrainable from above • Tuning = right mean answer for nonright reasons • Prospects for understanding cloud changes • in ensembles of runs of ensembles of GCMs • via conceptualizations • a lot is being learned as uncertainty fails to shrink
Elementary and rigorous • Maxwell’s equations • (from Robin Hogan, Reading U, ECMWF seminar 2008) • Now just integrate all energy over all matter! http://www.met.rdg.ac.uk/clouds/maxwell/ total “CRF”
From particles to continuum Maxwell E,B for . ensemble Key bulk variable: Extinction b (units: m-1) Robin Hogan ECMWF Seminar 2008
From particles to continuum Maxwell E,B for . ensemble • Bulk variable: Extinction b (units: m-1) • Shortwave: ~all scattering, ~no absorption • proportional to cross section (condensate volume/re) • re is “effective radius” (3rd moment/2nd moment of DSD) • Longwave: mostly absorption (& emission) • proportional to condensate volume (mass) • no re (droplet size) dependence! • typically ~2 times greater than SW scattering extinction
Angle integral Maxwell eqs. particle ensemble angle integral Robin Hogan ECMWF Seminar 2008
Wavelength integral can be done Maxwell eqs. particle ensemble angle integral wavelength integral • Complicated for gases but • Yields to precision laboratory (controlled) empiricism • leveraged with physics • Captured/ simplified in clever bundling • ‘bands’ of abs. coeff. k • Tuned up with final broadband empirical calibrations • clouds mercifully gray
“Unreasonable” assumptions Maxwell eqs. particle ensemble angle integral Robin Hogan ECMWF Seminar 2008
Unbiased vs. accurate • The vastness of our integral can be useful • don’t need the integrand accurate and complete • merely need a sufficiently large and unbiased sample, of an unbiased estimator of it! • Example: McICA radiation • Independent Column Approximation (ICA) • Monte Carlo (Mc) treatment of wavelength integral
Locally wrong, but unbiased • ICA: • Neglect hor. photon flux Fhor (3D effects) • Wrong alm. ev. in inhomogeneous clouds • But unbiased since • MC: • Send different wavelength bands through each subgrid cloud overlap realization • Unbiased, and large-enough subsample of vast 2D space • (even for weather forecasts)
Hooray for atmospheric physics! Maxwell eqs. z up to TOA (overlaps) seasons ENSO... small-meso structure geo-space (lat, lon) particle ensemble angle integral wavelength integral Nice solid rules and tools! (who remembers “anomalous absorption” ?) ∫∫∫∫∫∫∫ R(longlived GHGs, T, qv, baerosol, qcond, phase, re) Now for the problem of space-time integration... (x,y,z,t) (x,y,z,t) (x,y,z,t)
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs to be constrainable from above • Tuning = right mean answer for nonright reasons • Prospects for understanding cloud changes • in ensembles of runs of ensembles of GCMs • via conceptualizations • a lot is learned even as uncertainty fails to shrink
THE PROBLEM OF SCALES in practice, (x,y,t) is really (x,y,t,scales)
Almost without limit...but remember the independent column approximation!
For a cloudy column, 2 things matter: Emission temperature, and albedo the ISCCP 2D space for characterizing cloudy columns
net CRF in that space • Kubar et al. (2007)
Project any cloud population (joint histogram in this space), and sum to get total CRF...presto!
Can do for any set of cloudy pixels – like these ‘cloud population regimes’ • from a cluster analysis (aka self-organizing maps) of daily 5 degree joint histograms in the ISCCP 2-space
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs • Prospects for understanding cloud changes • in ensembles of runs of ensembles of GCMs • via conceptualizations • a lot is being learned even as uncertainty fails to shrink
Stuck with 3D models • The vastness of our integral can be a pain too • We only have laws to predict cloudiness in 3D • where air saturates, fundamentally • where air almost-saturates, for scale-truncated fluid dynamics/ thermodynamics • implying cloud in unresolved smaller-scale fluctuations • GCMs are stuck integrating partly-cloudy radiation over z
First off, cloud water is a precision nightmare • only a few % of the water is condensed 0.3 60 mm = kg/m2
Problems with sums and integrals • The radiative impact of a local volume of cloudiness is highly nonlinear • so it matters what’s above/below LW SW opacity t a condensate path/re
Magic number courtesy Robin Hogan
Yet all these effects are secondary! • (cloud fraction at each model level, condensed water at each model level) courtesy Robin Hogan
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs • tuning • Prospects for understanding cloud changes • in ensembles of runs of ensembles of GCMs • in observations • via conceptualizations • a lot is being learned even as uncertainty fails to shrink
Tuning • GCM cloudy radiation is tuned • by several or 10s of Watts, I think • to have net flux =0 for preindustrial control climate • in each latitude belt • to have right SW and LW individually? (somebody correct me if wrong?) Does this constrain sensitivity? no such luck
Back to the integral -- tolerances Global warming is driven by the imbalance [∫∫∫∫∫∫swR- ∫∫∫∫∫∫LwR] <1 Wm-2out of 235. Hansen et al. 2004 Science Express
Outline • Preamble: clouds as a climate feedback • A step backward: stating the problem flatly • Triumphs of atm. RT column physics • Now about cloudiness (x,y,z,t)... • Statistical descriptions from observations • Formulating GCMs to be constrainable from above • Tuning • Prospects for understanding cloud changes • in models • in obs • via conceptualizations • a lot is being learned even as uncertainty fails to shrink
What could change systematically with climate? • Large-scale cloud coverage? • say low cloud incr. due to static stability increase? • Miller 1997 negative feedback • but see Wood-Bretherton 2006 EIS recasting